01Safety
0.0 / 30
What changed in the harness
Selection accuracy 64→69, token cost up 3%, unconfirmed writes 0%→0%.
Category breakdown
Where the score comes from.
Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.
02Legibility
22.9 / 30
03Economics
20.0 / 20
04Discoverability
11.1 / 20
Highest-impact fix
Estimated gain +30 pointsAdd explicit identity and permission preflight tools
Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.
Description evidence
Defects and rewrites.
2 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.
| Tool | Defect types | Suggested rewrite |
|---|---|---|
read_urls |
no_return_description |
Read up to 20 public web pages in one batch call, returning each page as a clean AIDocument: Markdown plus title, description, and structure, with navigation and boilerplate stripped. Faster than calling read_url repeatedly; use it to compare or summarize multiple pages at once. A failed URL is reported per-item and does not block the others. Powered by Lyrenth. |
check_usage |
no_return_description |
Return your Lyrenth credit usage as a report: plan tier, credits used against your monthly limit, credits remaining, and the reset date. Takes no arguments. |
Selection evidence
Confusable tool pairs.
1 pair where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
read_url |
read_urls |
low | Descriptions and schemas strongly disambiguate by count: read_url takes a single string url, read_urls takes an array of 1-20. Natural-language tasks almost always signal singular ('read this page...') or plural ('read/compare these pages...'), mapping cleanly. Genuine confusion would only arise for vague tasks without a stated URL count (e.g., 'get the content of that site'), where the agent might pick the singular tool and under-deliver — but this is a weak stretch, not a real ambiguity. |
Compare the field